# load train and test data
print('loading data ...')
# (x_train, y_train), (x_test, y_test) = Funcs.load_data(dirs, target_size, gray=gray, classes=classes)
(x_train, y_train), (x_test, y_test) = load_data()
(x_train, y_train) = Funcs.trans_data(x_train, y_train, classes)
(x_test, y_test) = Funcs.trans_data(x_test, y_test, classes)
print('loading data done')
# define callbacks for models to render model performance better
checkpoint = ModelCheckpoint(model_save_path,
monitor='val_acc',
save_best_only=True,
verbose=0)
lrschedual = LearningRateScheduler(
lambda epoch: Funcs.lr_schedual(epoch, epochs=epochs), verbose=0)
# whether load existent model
if load_existent_model and os.path.exists(model_load_path):
model = load_model(model_load_path)
else:
model = model_design(x_train.shape[1:], classes)
# train the model
start_time = datetime.datetime.now()
model.fit(x_train,
y_train,
epochs=epochs,
batch_size=batch_size,
validation_data=(x_test, y_test),
callbacks=[checkpoint, lrschedual],
gray = False
load_existent_model = False
# load train and test data
print('loading data ...')
(images, labels, counts) = Funcs.read_images_from_directories(dirs, target_size, get_counts=True, init_json=True,
shuffle=True)
classes = len(counts)
(x_train, y_train), (x_test, y_test) = Funcs.split_train_test(images, labels, counts)
(x_train, y_train) = Funcs.trans_data(x_train, y_train, classes)
(x_test, y_test) = Funcs.trans_data(x_test, y_test, classes)
print('loading data done')
# define callbacks for models to render model performance better
checkpoint = ModelCheckpoint(model_save_path, monitor='val_acc', save_best_only=True, verbose=0)
lrschedual = LearningRateScheduler(lambda epoch: Funcs.lr_schedual(epoch, epochs=epochs), verbose=0)
# whether load existent model
if load_existent_model and os.path.exists(model_load_path):
model = load_model(model_load_path)
else:
model = model_design(x_train.shape[1:], classes)
# train the model
start_time = datetime.datetime.now()
model.fit(x_train, y_train, epochs=epochs, batch_size=batch_size,
validation_data=(x_test, y_test), callbacks=[checkpoint, lrschedual], verbose=1)
# fine tune last conv parameters
# for layer in model.layers[:15]:
# layer.trainable = False